Article

The Weight of One

What happens when one person holds so much sway over an AI company that a crisis at the top can force customers, suppliers, and investors to rethink their own plans?

Oscar Scarano Week 27 Leer en español
Share Share on LinkedIn Share on X Share on Facebook Share by email
abstract person at the center of floating structure
AI assisted/generated image

A chip does not become slower because its chief executive loses the confidence of a board. Software does not stop running because a founder leaves. Yet a leadership crisis can alter the assumptions behind investments, partnerships and plans far beyond the company where it began.

On November 12, Nvidia announced Jensen Huang’s resignation and appointed chief financial officer Colette Kress interim CEO. His departure followed an investigation announced four days earlier into reports that he had been targeted by a foreign intelligence effort. The board had said it had no reason to believe national security or company secrets were compromised. China denied the espionage allegations. A leadership change does not settle those allegations.

The episode exposes a problem that will outlast its headlines: the economic weight placed on a person can grow much faster than the institutions built to manage that dependence.

How a Person Becomes Infrastructure

Exceptional individuals have carried enormous economic influence before. During the Panic of 1907, J. P. Morgan and other financiers played a central role in coordinating the response. The crisis helped drive the movement that eventually produced the Federal Reserve. Dependence on a few powerful people is an old problem.

In AI, that dependence can connect several layers of business at once. A founder may help set the product roadmap, negotiate major partnerships, attract capital and maintain relationships with governments. Customers can end up relying on both the technology and the person who makes commitments about its future.

This distinction matters. A company can employ thousands of talented people while leaving too many consequential relationships dependent on one individual. Its technical capacity may be distributed; confidence in its direction may be much more concentrated.

Huang, Elon Musk and Dario Amodei should not be treated as interchangeable measures of economic power. The useful comparison concerns the decisions concentrated around leaders, the businesses exposed to those decisions and the arrangements that would survive their departure.

There is already a concrete precedent for placing limits on an executive’s discretion. Musk’s 2018 “funding secured” statements led to SEC enforcement and a settlement requiring oversight of certain communications. The underlying issue was the effect of a leader’s statements on investors and the adequacy of company controls.

A company’s ability to withstand the loss of its founder should be part of what makes it valuable.

The Shock Travels Through Expectations

A fall in market capitalization is not the same as cash disappearing from a company’s bank account. Nor does it prove that customers have cancelled orders or moved to competitors. Stock prices can change far faster than supply chains.

For an AI customer, replacing a major technology supplier may require changing software, retraining staff, securing capacity and accepting new operating costs. Those difficulties can protect a supplier’s business even while investors reassess its leadership. They also leave customers exposed to decisions they cannot easily escape.

The FTC’s examination of major cloud–AI partnerships identified potential restrictions on access to computing inputs and increased contractual and technical switching costs. The Financial Stability Board has separately identified provider concentration and third-party dependencies as potential vulnerabilities in finance. These findings concern specific markets, but they offer a useful way to examine dependence across the AI economy.

Patents are only one part of that picture. Software compatibility, access to computing capacity, accumulated expertise and contractual commitments can also make alternatives difficult to use. Publishing code alone would not remove every dependency.

A leadership shock could therefore travel through expectations before affecting operations. Investors might reassess funding. A new management team might reconsider commitments. Customers might delay expansion while seeking clarity. Suppliers could adjust their own plans in response.

These are possible transmission channels, not an assertion that every one will occur at Nvidia. Their importance depends on the actual contracts, financing and alternatives involved.

The same concentration that helped an industry expand quickly can make a change of direction costly for many participants at once.

Responsibility Without a Permit for Every Decision

Personal conduct becomes a legitimate corporate concern when it creates a conflict of interest, compromises security, involves abuse of authority or exposes others to material undisclosed risks. A powerful executive should retain a private life. A company should also have procedures for the points where private relationships intersect with professional obligations.

Those procedures need to work when applied to the founder.

Independent oversight, clear disclosure duties, delegated authority and credible succession arrangements are practical places to begin. For customers of essential services, continuity planning and workable migration options deserve attention too.

Some of this is ordinary board responsibility. Some may require regulation where a company’s decisions impose substantial risks beyond its shareholders. The case for intervention should identify that wider exposure and explain why existing arrangements cannot adequately address it.

Regulation has costs. Complex approvals can delay investment, and fixed compliance expenses can make life harder for smaller competitors. OECD research finds that anticompetitive regulatory barriers can weaken productivity. That is a reason to scrutinize the design of rules, rather than assume that their number tells us their economic effect.

A safeguard should earn its place by reducing a defined risk at a reasonable cost. Shared reporting standards may help; duplicative paperwork may simply consume resources. The OECD’s voluntary Hiroshima AI reporting framework offers one existing mechanism for comparing organizational governance practices across jurisdictions. It does not amount to global supervision of executives.

International coordination would also face a difficult incentive problem: governments want accountable companies, but they want successful national champions.

Private Firms, National Priorities, Global Exposure

The state already helps shape the AI market. Export controls determine which customers can obtain certain advanced chips. In January 2026, the US revised its licensing policy to allow case-by-case review of exports of Nvidia H200 and comparable processors to China under specified security conditions. Access to a major market was a public-policy decision as well as a commercial ambition. 

This relationship creates pressure in both directions. Companies seek favorable rules and access to customers. Governments seek investment, technological advantage and security. A government can fear excessive corporate power while also fearing the consequences of weakening a strategically important company.

National-security controls can impose real commercial costs. They can also pursue objectives that a company’s sales figures cannot measure. Equally, protecting a national champion can preserve capabilities while entrenching dependence on it.

A public alternative would need its own scrutiny. State ownership does not by itself ensure competent management, open access or continuity. Political decisions can also concentrate power and transmit shocks.

The practical test should apply across ownership models: who makes consequential decisions, who can challenge them, who bears the costs and what alternatives remain available?

That test is particularly important when customers and investors are global but corporate oversight and security policy remain largely national.

Building for the Person Who Leaves

For boards, the useful exercise is to examine what would happen if the founder could no longer lead tomorrow. Which commitments are documented? Which relationships have another responsible executive? Which decisions can proceed? For customers, the equivalent exercise is to understand how long a supplier change would take and what it would cost.

Those questions matter whether a departure follows misconduct, illness, disagreement or retirement. Exceptional leadership and institutional resilience should develop together.

Huang’s departure will test Nvidia’s arrangements. It also gives the wider industry a reason to examine its own. Building powerful machines does not eliminate human fallibility. The economic consequences depend partly on how much authority, trust and dependence have accumulated around the person who falters.

An industry’s maturity is measured partly by what it can keep doing when its most influential person leaves the room.

LinkedIn

Continue the conversation on LinkedIn

LinkedIn

More to read

Business Designing for Who Watches Culture The Museum of What’s Missing Article Ask Before You Quote

Continue Reading

Explore topics

AI & Society Automation Business Culture Design Human Work +